Developing Story
LLMs as Oracles: Reliance on AI for Personal Decisions
New research quantifies the rising trend of people using LLMs as 'oracles' for subjective personal decisions, finding this behavior has grown since 2023 and skews younger. The findings feed a broader narrative about AI's psychological influence and inform emerging litigation and regulatory debates around chatbot safety design.
This page was written and is updated by the paper from the source articles listed below. Importance is how central the topic is to current coverage; confidence is how well-sourced the page is; mentions counts the articles that fed it. How the wiki works →
## Overview
Researchers have identified and quantified a growing pattern of "LLM-as-oracle" use, where people treat AI chatbots as all-knowing authorities for subjective personal questions — offloading judgment and decision-making that was previously the domain of friends, family, therapists, or personal reflection (arXiv:2609.14849, September 15).
## Key Findings
Using a new typology and LLM-based measurement methods applied to 68,000 prompts drawn from public usage datasets (WildChat and ThoughtTrace), researchers found that oracle-style reliance on LLMs has increased steadily from 2023 to 2026 and is disproportionately prevalent among younger users (arXiv:2609.14849, September 15). The paper frames this as a risk to user autonomy and psychological well-being, since subjective personal questions (relationship advice, moral dilemmas, life decisions) may not have objectively correct answers that an AI model can reliably provide.
## Why This Matters
This research provides an evidentiary and methodological foundation for a wider narrative about AI's psychological and social influence — overlapping with concerns about teen overreliance on companion chatbots, AI's role in high-stakes personal advice (e.g., FaithfulBench's findings on religious counsel), and broader debates about AI safety in emotionally sensitive use cases.
## Strategic Relevance
For litigators and regulators, this data offers a basis to argue that general-purpose chatbots function as de facto advice-givers without appropriate safeguards, informing potential claims around inadequate disclaimers, deceptive design, or duty-of-care standards. For AI companies, the trend signals reputational and regulatory risk requiring guardrails (e.g., prompts nudging users toward professional help, disclaimers on subjective questions) as usage of AI for life-guidance queries continues to climb.
## Related Developments
This connects to broader industry-wide reckoning with chatbot overreliance, including OpenAI's ChatGPT mass tort litigation, growing use of age assurance features, and research on faith-based counsel consistency (FaithfulBench). It also intersects with ongoing debate about whether frontier AI labs are doing enough to mitigate psychological dependency risks as usage scales.